1 Updating Information for FreeBSD current users
3 This file is maintained and copyrighted by M. Warner Losh <imp@freebsd.org>.
4 See end of file for further details. For commonly done items, please see the
5 COMMON ITEMS: section later in the file. These instructions assume that you
6 basically know what you are doing. If not, then please consult the FreeBSD
9 http://www.freebsd.org/doc/en_US.ISO8859-1/books/handbook/makeworld.html
11 Items affecting the ports and packages system can be found in
12 /usr/ports/UPDATING. Please read that file before running portupgrade.
15 The latest version of openssl rejects TLS handshakes with DH
16 parameters below 768 bits. sendmail releases prior to 8.15.2
17 (not yet released), defaulted to a 512 bit DH parameter setting
18 for client connections. To improve interoperability, the
19 sendmail default for client connections has been raised to
23 It is no longer necessary to include the dwarf version in your DEBUG
24 options in your kernel config file. The bug that required it to be
25 placed in the config file has bene fixed. DEBUG should now just
26 contain -g. The build system will automatically update things
27 to do the right thing.
33 On i386 and amd64 systems, the onifconsole flag is now set by default
34 in /etc/ttys for ttyu0. This causes ttyu0 to be automatically enabled
35 as a login TTY if it is set in the bootloader as an active kernel
36 console. No changes in behavior should result otherwise. To revert to
37 the previous behavior, set ttyu0 to "off" in /etc/ttys.
40 Clang and llvm have been upgraded to 3.4.1 release.
43 Clang and llvm have been upgraded to 3.4 release.
46 The nve(4) driver for NVIDIA nForce MCP Ethernet adapters has
47 been deprecated and will not be part of FreeBSD 11.0 and later
48 releases. If you use this driver, please consider switching to
49 the nfe(4) driver instead.
52 The behavior of gss_pseudo_random() for the krb5 mechanism
53 has changed, for applications requesting a longer random string
54 than produced by the underlying enctype's pseudo-random() function.
55 In particular, the random string produced from a session key of
56 enctype aes256-cts-hmac-sha1-96 or aes256-cts-hmac-sha1-96 will
57 be different at the 17th octet and later, after this change.
58 The counter used in the PRF+ construction is now encoded as a
59 big-endian integer in accordance with RFC 4402.
60 __FreeBSD_version is bumped to 902505.
66 Behavior of devfs rules path matching has been changed.
67 Pattern is now always matched against fully qualified devfs
68 path and slash characters must be explicitly matched by
69 slashes in pattern (FNM_PATHNAME). Rulesets involving devfs
70 subdirectories must be reviewed.
73 hastctl(8)'s `status' command output changed to terse one-liner format.
74 Scripts using this should switch to `list' command or be rewritten.
77 Fix a bug that allowed a tracing process (e.g. gdb) to write
78 to a memory-mapped file in the traced process's address space
79 even if neither the traced process nor the tracing process had
80 write access to that file.
83 Added ZFS TRIM support which is enabled by default. To disable
84 ZFS TRIM support set vfs.zfs.trim.enabled=0 in loader.conf.
86 Creating new ZFS pools and adding new devices to existing pools
87 first performs a full device level TRIM which can take a significant
88 amount of time. The sysctl vfs.zfs.vdev.trim_on_init can be set to 0
89 to disable this behaviour.
91 ZFS TRIM requires the underlying device support BIO_DELETE which
92 is currently provided by methods such as ATA TRIM and SCSI UNMAP
93 via CAM, which are typically supported by SSD's.
95 Stats for ZFS TRIM can be monitored by looking at the sysctl's
96 under kstat.zfs.misc.zio_trim.
99 `list' command has been added to hastctl(8). For now, it is full
100 equivalent of `status' command.
101 WARNING: in the near future the output of hastctl's status command
102 will change to more terse format. If you use `hastctl status'
103 for parsing in your scripts, switch to `hastctl list'.
106 The mergemaster command now uses the default MAKEOBJDIRPREFIX
107 rather than creating it's own in the temporary directory in
108 order allow access to bootstrapped versions of tools such as
109 install and mtree. When upgrading from version of FreeBSD where
110 the install command does not support -l, you will need to
111 install a new mergemaster command if mergemaster -p is required.
112 This can be accomplished with the command (cd src/usr.sbin/mergemaster
115 Due to the use of the new -l option to install(1) during build
116 and install, you must take care not to directly set the INSTALL
117 make variable in your /etc/make.conf, /etc/src.conf, or on the
118 command line. If you wish to use the -C flag for all installs
119 you may be able to add INSTALL+=-C to /etc/make.conf or
123 Fix a bug that allows NFS clients to issue READDIR on files.
126 The install(1) option -M has changed meaning and now takes an
127 argument that is a file or path to append logs to. In the
128 unlikely event that -M was the last option on the command line
129 and the command line contained at least two files and a target
130 directory the first file will have logs appended to it. The -M
131 option served little practical purpose in the last decade so it's
132 used expected to be extremely rare.
135 A new compression method (lz4) has been merged to. Please refer to
136 zpool-features(7) for more information.
138 Please refer to the "ZFS notes" section of this file for information
139 on upgrading boot ZFS pools.
142 The VFS KBI was changed with the merge of several nullfs
143 optimizations and fixes. All filesystem modules must be
147 With the addition of auditdistd(8), a new auditdistd user is now
148 depended on during installworld. "mergemaster -p" can be used to add
149 the user prior to installworld, as documented in the handbook.
155 A new version of ZFS (pool version 5000) has been merged to 9-STABLE.
156 Starting with this version the old system of ZFS pool versioning
157 is superseded by "feature flags". This concept enables forward
158 compatibility against certain future changes in functionality of ZFS
159 pools. The first two read-only compatible "feature flags" for ZFS
160 pools are "com.delphix:async_destroy" and "com.delphix:empty_bpobj".
161 For more information read the new zpool-features(7) manual page.
162 Please refer to the "ZFS notes" section of this file for information
163 on upgrading boot ZFS pools.
166 The commit introducing bsd.compiler.mk breaks the traditional
167 building of kernels before this point. Add -m ${SRC}/share/mk
168 (for the right value of SRC) to your command lines to work
169 around; update your useland to a point after this; or use the
170 buildkernel/installkernel top-level targets. See also 20120829.
173 The IPFIREWALL_FORWARD kernel option has been removed. Its
174 functionality now turned on by default.
177 The random(4) support for the VIA hardware random number
178 generator (`PADLOCK') is no longer enabled unconditionally.
179 Add the PADLOCK_RNG option in the custom kernel config if
180 needed. The GENERIC kernels on i386 and amd64 do include the
181 option, so the change only affects the custom kernel
185 The amd64 kernel now uses xsetbv, xrstor instructions. To compile with
186 the traditional method, you must update your system with an installworld
187 before the kernel will build. The documented make buildkernel/installkernel
188 interfaces (coupled with fresh make kernel-toolchain) continue to work.
191 The sparc64 ZFS loader has been changed to no longer try to auto-
192 detect ZFS providers based on diskN aliases but now requires these
193 to be explicitly listed in the OFW boot-device environment variable.
196 Now unix domain sockets behave "as expected" on nullfs(5). Previously
197 nullfs(5) did not pass through all behaviours to the underlying layer,
198 as a result if we bound to a socket on the lower layer we could connect
199 only to the lower path; if we bound to the upper layer we could connect
200 only to the upper path. The new behavior is one can connect to both the
201 lower and the upper paths regardless what layer path one binds to.
204 The acpi_wmi(4) status device /dev/wmistat has been renamed to
208 A new VOP_ADVISE() was added to support posix_fadvise(2). All
209 filesystem modules must be recompiled.
212 The interface of the VOP_VPTOCNP(9) changed, now the returned
213 vnode shall be referenced, previously it was required to be
214 only held. All in-tree filesystems are converted.
220 The broken amd(4) driver has been replaced with esp(4) in the amd64,
221 i386 and pc98 GENERIC kernel configuration files.
224 This commit modifies vfs_register() so that it uses a hash
225 calculation to set vfc_typenum, which is enabled by default.
226 The first time a system is booted after this change, the
227 vfc_typenum values will change for all file systems. The
228 main effect of this is a change to the NFS server file handles
229 for file systems that use vfc_typenum in their fsid, such as ZFS.
230 It will, however, prevent vfc_typenum from changing when file
231 systems are loaded in a different order for subsequent reboots.
232 To disable this, you can set vfs.typenumhash=0 in /boot/loader.conf
233 until you are ready to remount all NFS clients after a reboot.
236 Bump the shared library version numbers for libraries that
237 do not use symbol versioning, have changed the ABI compared
238 to stable/8 and which shared library version was not bumped.
239 Done as part of 9.0-RELEASE cycle.
242 During the merge of Capsicum features, the fget(9) KPI was modified.
243 This may require the rebuilding of out-of-tree device drivers --
244 issues have been reported specifically with the nVidia device driver.
245 __FreeBSD_version is bumped to 900041.
247 Also, there is a period between 20110811 and 20110814 where the
248 special devices /dev/{stdin,stdout,stderr} did not work correctly.
249 Building world from a kernel during that window may not work.
252 The packet filter (pf) code has been updated to OpenBSD 4.5.
253 You need to update userland tools to be in sync with kernel.
254 This update breaks backward compatibility with earlier pfsync(4)
255 versions. Care must be taken when updating redundant firewall setups.
258 The following sysctls and tunables are retired on x86 platforms:
260 machdep.hlt_logical_cpus
261 The following sysctl is retired:
262 machdep.hyperthreading_allowed
263 The sysctls were supposed to provide a way to dynamically offline and
264 online selected CPUs on x86 platforms, but the implementation has not
265 been reliable especially with SCHED_ULE scheduler.
266 machdep.hyperthreading_allowed tunable is still available to ignore
267 hyperthreading CPUs at OS level.
268 Individual CPUs can be disabled using hint.lapic.X.disabled tunable,
269 where X is an APIC ID of a CPU. Be advised, though, that disabling
270 CPUs in non-uniform fashion will result in non-uniform topology and
271 may lead to sub-optimal system performance with SCHED_ULE, which is
275 cpumask_t type is retired and cpuset_t is used in order to describe
279 Changes to ifconfig(8) for dynamic address family detection mandate
280 that you are running a kernel of 20110525 or later. Make sure to
281 follow the update procedure to boot a new kernel before installing
285 Support for sun4v architecture is officially dropped
288 Several KPI breaking changes have been committed to the mii(4) layer,
289 the PHY drivers and consequently some Ethernet drivers using mii(4).
290 This means that miibus.ko and the modules of the affected Ethernet
291 drivers need to be recompiled.
293 Note to kernel developers: Given that the OUI bit reversion problem
294 was fixed as part of these changes all mii(4) commits related to OUIs,
295 i.e. to sys/dev/mii/miidevs, PHY driver probing and vendor specific
296 handling, no longer can be merged verbatim to stable/8 and previous
300 Users of the Atheros AR71xx SoC code now need to add 'device ar71xx_pci'
301 into their kernel configurations along with 'device pci'.
304 The default NFS client is now the new NFS client, so fstype "newnfs"
305 is now "nfs" and the regular/old NFS client is now fstype "oldnfs".
306 Although mounts via fstype "nfs" will usually work without userland
307 changes, it is recommended that the mount(8) and mount_nfs(8)
308 commands be rebuilt from sources and that a link to mount_nfs called
309 mount_oldnfs be created. The new client is compiled into the
310 kernel with "options NFSCL" and this is needed for diskless root
311 file systems. The GENERIC kernel configs have been changed to use
312 NFSCL and NFSD (the new server) instead of NFSCLIENT and NFSSERVER.
313 To use the regular/old client, you can "mount -t oldnfs ...". For
314 a diskless root file system, you must also include a line like:
316 vfs.root.mountfrom="oldnfs:"
318 in the boot/loader.conf on the root fs on the NFS server to make
319 a diskless root fs use the old client.
322 The GENERIC kernels for all architectures now default to the new
323 CAM-based ATA stack. It means that all legacy ATA drivers were
324 removed and replaced by respective CAM drivers. If you are using
325 ATA device names in /etc/fstab or other places, make sure to update
326 them respectively (adX -> adaY, acdX -> cdY, afdX -> daY, astX -> saY,
327 where 'Y's are the sequential numbers starting from zero for each type
328 in order of detection, unless configured otherwise with tunables,
329 see cam(4)). There will be symbolic links created in /dev/ to map
330 old adX devices to the respective adaY. They should provide basic
331 compatibility for file systems mounting in most cases, but they do
332 not support old user-level APIs and do not have respective providers
333 in GEOM. Consider using updated management tools with new device names.
335 It is possible to load devices ahci, ata, siis and mvs as modules,
336 but option ATA_CAM should remain in kernel configuration to make ata
337 module work as CAM driver supporting legacy ATA controllers. Device ata
338 still can be used in modular fashion (atacore + ...). Modules atadisk
339 and atapi* are not used and won't affect operation in ATA_CAM mode.
340 Note that to use CAM-based ATA kernel should include CAM devices
341 scbus, pass, da (or explicitly ada), cd and optionally others. All of
342 them are parts of the cam module.
344 ataraid(4) functionality is now supported by the RAID GEOM class.
345 To use it you can load geom_raid kernel module and use graid(8) tool
346 for management. Instead of /dev/arX device names, use /dev/raid/rX.
348 No kernel config options or code have been removed, so if a problem
349 arises, please report it and optionally revert to the old ATA stack.
350 In order to do it you can remove from the kernel config:
355 , and instead add back:
356 device atadisk # ATA disk drives
357 device ataraid # ATA RAID drives
358 device atapicd # ATAPI CDROM drives
359 device atapifd # ATAPI floppy drives
360 device atapist # ATAPI tape drives
363 The default NFS server has been changed to the new server, which
364 was referred to as the experimental server. If you need to switch
365 back to the old NFS server, you must now put the "-o" option on
366 both the mountd and nfsd commands. This can be done using the
367 mountd_flags and nfs_server_flags rc.conf variables until an
368 update to the rc scripts is committed, which is coming soon.
371 The GNU Objective-C runtime library (libobjc), and other Objective-C
372 related components have been removed from the base system. If you
373 require an Objective-C library, please use one of the available ports.
376 ath(4) has been split into bus- and device- modules. if_ath contains
377 the HAL, the TX rate control and the network device code. if_ath_pci
378 contains the PCI bus glue. For Atheros MIPS embedded systems, if_ath_ahb
379 contains the AHB glue. Users need to load both if_ath_pci and if_ath
380 in order to use ath on everything else.
382 TO REPEAT: if_ath_ahb is not needed for normal users. Normal users only
383 need to load if_ath and if_ath_pci for ath(4) operation.
386 As part of the replacement of sysinstall, the process of building
387 release media has changed significantly. For details, please re-read
388 release(7), which has been updated to reflect the new build process.
391 GNU binutils 2.17.50 (as of 2007-07-03) has been merged to -HEAD. This
392 is the last available version under GPLv2. It brings a number of new
393 features, such as support for newer x86 CPU's (with SSE-3, SSSE-3, SSE
394 4.1 and SSE 4.2), better support for powerpc64, a number of new
395 directives, and lots of other small improvements. See the ChangeLog
396 file in contrib/binutils for the full details.
399 IPsec's HMAC_SHA256-512 support has been fixed to be RFC4868
400 compliant, and will now use half of hash for authentication.
401 This will break interoperability with all stacks (including all
402 actual FreeBSD versions) who implement
403 draft-ietf-ipsec-ciph-sha-256-00 (they use 96 bits of hash for
405 The only workaround with such peers is to use another HMAC
406 algorithm for IPsec ("phase 2") authentication.
409 Remove the uio_yield prototype and symbol. This function has
410 been misnamed since it was introduced and should not be
411 globally exposed with this name. The equivalent functionality
412 is now available using kern_yield(curthread->td_user_pri).
413 The function remains undocumented.
416 A SYSCTL_[ADD_]UQUAD was added for unsigned uint64_t pointers,
417 symmetric with the existing SYSCTL_[ADD_]QUAD. Type checking
418 for scalar sysctls is defined but disabled. Code that needs
419 UQUAD to pass the type checking that must compile on older
420 systems where the define is not present can check against
421 __FreeBSD_version >= 900030.
423 The system dialog(1) has been replaced with a new version previously
424 in ports as devel/cdialog. dialog(1) is mostly command-line compatible
425 with the previous version, but the libdialog associated with it has
426 a largely incompatible API. As such, the original version of libdialog
427 will be kept temporarily as libodialog, until its base system consumers
428 are replaced or updated. Bump __FreeBSD_version to 900030.
431 If you are trying to run make universe on a -stable system, and you get
432 the following warning:
433 "Makefile", line 356: "Target architecture for i386/conf/GENERIC
434 unknown. config(8) likely too old."
435 or something similar to it, then you must upgrade your -stable system
436 to 8.2-Release or newer (really, any time after r210146 7/15/2010 in
437 stable/8) or build the config from the latest stable/8 branch and
438 install it on your system.
440 Prior to this date, building a current universe on 8-stable system from
441 between 7/15/2010 and 1/2/2011 would result in a weird shell parsing
442 error in the first kernel build phase. A new config on those old
443 systems will fix that problem for older versions of -current.
446 The TCP stack has been modified to allow Khelp modules to interact with
447 it via helper hook points and store per-connection data in the TCP
448 control block. Bump __FreeBSD_version to 900029. User space tools that
449 rely on the size of struct tcpcb in tcp_var.h (e.g. sockstat) need to
453 Generic IEEE 802.3 annex 31B full duplex flow control support has been
454 added to mii(4) and bge(4), bce(4), msk(4), nfe(4) and stge(4) along
455 with brgphy(4), e1000phy(4) as well as ip1000phy() have been converted
456 to take advantage of it instead of using custom implementations. This
457 means that these drivers now no longer unconditionally advertise
458 support for flow control but only do so if flow control is a selected
459 media option. This was implemented in the generic support that way in
460 order to allow flow control to be switched on and off via ifconfig(8)
461 with the PHY specific default to typically off in order to protect
462 from unwanted effects. Consequently, if you used flow control with
463 one of the above mentioned drivers you now need to explicitly enable
465 ifconfig bge0 media auto mediaopt flowcontrol
467 Along with the above mentioned changes generic support for setting
468 1000baseT master mode also has been added and brgphy(4), ciphy(4),
469 e1000phy(4) as well as ip1000phy(4) have been converted to take
470 advantage of it. This means that these drivers now no longer take the
471 link0 parameter for selecting master mode but the master media option
472 has to be used instead, for example like in the following:
473 ifconfig bge0 media 1000baseT mediaopt full-duplex,master
475 Selection of master mode now is also available with all other PHY
476 drivers supporting 1000baseT.
479 The TCP stack has received a significant update to add support for
480 modularised congestion control and generally improve the clarity of
481 congestion control decisions. Bump __FreeBSD_version to 900025. User
482 space tools that rely on the size of struct tcpcb in tcp_var.h (e.g.
483 sockstat) need to be recompiled.
486 The man(1) utility has been replaced by a new version that no longer
487 uses /etc/manpath.config. Please consult man.conf(5) for how to
488 migrate local entries to the new format.
491 The copyright strings printed by login(1) and sshd(8) at the time of a
492 new connection have been removed to follow other operating systems and
496 A workaround for a fixed ld bug has been removed in kernel code,
497 so make sure that your system ld is built from sources after
498 revision 210245 from 2010-07-19 (r211583 if building head kernel
499 on stable/8, r211584 for stable/7; both from 2010-08-21).
500 A symptom of incorrect ld version is different addresses for
501 set_pcpu section and __start_set_pcpu symbol in kernel and/or modules.
504 The $ipv6_prefer variable in rc.conf(5) has been split into
505 $ip6addrctl_policy and $ipv6_activate_all_interfaces.
507 The $ip6addrctl_policy is a variable to choose a pre-defined
508 address selection policy set by ip6addrctl(8). A value
509 "ipv4_prefer", "ipv6_prefer" or "AUTO" can be specified. The
512 The $ipv6_activate_all_interfaces specifies whether IFDISABLED
513 flag (see an entry of 20090926) is set on an interface with no
514 corresponding $ifconfig_IF_ipv6 line. The default is "NO" for
515 security reason. If you want IPv6 link-local address on all
516 interfaces by default, set this to "YES".
518 The old ipv6_prefer="YES" is equivalent to
519 ipv6_activate_all_interfaces="YES" and
520 ip6addrctl_policy="ipv6_prefer".
523 DTrace has grown support for userland tracing. Due to this, DTrace is
524 now i386 and amd64 only.
525 dtruss(1) is now installed by default on those systems and a new
526 kernel module is needed for userland tracing: fasttrap.
527 No changes to your kernel config file are necessary to enable
528 userland tracing, but you might consider adding 'STRIP=' and
529 'CFLAGS+=-fno-omit-frame-pointer' to your make.conf if you want
530 to have informative userland stack traces in DTrace (ustack).
533 The acpi_aiboost(4) driver has been removed in favor of the new
534 aibs(4) driver. You should update your kernel configuration file.
537 BSD grep has been imported to the base system and it is built by
538 default. It is completely BSD licensed, highly GNU-compatible, uses
539 less memory than its GNU counterpart and has a small codebase.
540 However, it is slower than its GNU counterpart, which is mostly
541 noticeable for larger searches, for smaller ones it is measurable
542 but not significant. The reason is complex, the most important factor
543 is that we lack a modern and efficient regex library and GNU
544 overcomes this by optimizing the searches internally. Future work
545 on improving the regex performance is planned, for the meantime,
546 users that need better performance, can build GNU grep instead by
547 setting the WITH_GNU_GREP knob.
550 Due to the import of powerpc64 support, all existing powerpc kernel
551 configuration files must be updated with a machine directive like this:
552 machine powerpc powerpc
554 In addition, an updated config(8) is required to build powerpc kernels
558 A new version of ZFS (version 15) has been merged to -HEAD.
559 This version uses a python library for the following subcommands:
560 zfs allow, zfs unallow, zfs groupspace, zfs userspace.
561 For full functionality of these commands the following port must
562 be installed: sysutils/py-zfs
565 'vm_page's are now hashed by physical address to an array of mutexes.
566 Currently this is only used to serialize access to hold_count. Over
567 time the page queue mutex will be peeled away. This changes the size
568 of pmap on every architecture. And requires all callers of vm_page_hold
569 and vm_page_unhold to be updated.
572 WITH_CTF can now be specified in src.conf (not recommended, there
573 are some problems with static executables), make.conf (would also
574 affect ports which do not use GNU make and do not override the
575 compile targets) or in the kernel config (via "makeoptions
577 When WITH_CTF was specified there before this was silently ignored,
578 so make sure that WITH_CTF is not used in places which could lead
579 to unwanted behavior.
582 The kernel option COMPAT_IA32 has been replaced with COMPAT_FREEBSD32
583 to allow 32-bit compatibility on non-x86 platforms. All kernel
584 configurations on amd64 and ia64 platforms using these options must
585 be modified accordingly.
588 The utmp user accounting database has been replaced with utmpx,
589 the user accounting interface standardized by POSIX.
590 Unfortunately the semantics of utmp and utmpx don't match,
591 making it practically impossible to support both interfaces.
592 The user accounting database is used by tools like finger(1),
593 last(1), talk(1), w(1) and ac(8).
595 All applications in the base system use utmpx. This means only
596 local binaries (e.g. from the ports tree) may still use these
597 utmp database files. These applications must be rebuilt to make
600 After the system has been upgraded, it is safe to remove the old
601 log files (/var/run/utmp, /var/log/lastlog and /var/log/wtmp*),
602 assuming their contents is of no importance anymore. Old wtmp
603 databases can only be used by last(1) and ac(8) after they have
604 been converted to the new format using wtmpcvt(1).
607 Introduce the kernel thread "deadlock resolver" (which can be enabled
608 via the DEADLKRES option, see NOTES for more details) and the
609 sleepq_type() function for sleepqueues.
612 The rc.firewall and rc.firewall6 were unified, and
613 rc.firewall6 and rc.d/ip6fw were removed.
614 According to the removal of rc.d/ip6fw, ipv6_firewall_* rc
615 variables are obsoleted. Instead, the following new rc
616 variables are added to rc.d/ipfw:
618 firewall_client_net_ipv6, firewall_simple_iif_ipv6,
619 firewall_simple_inet_ipv6, firewall_simple_oif_ipv6,
620 firewall_simple_onet_ipv6, firewall_trusted_ipv6
622 The meanings correspond to the relevant IPv4 variables.
628 The default terminal emulation for syscons(4) has been changed
629 from cons25 to xterm on all platforms except pc98. This means
630 that the /etc/ttys file needs to be updated to ensure correct
631 operation of applications on the console.
633 The terminal emulation style can be toggled per window by using
634 vidcontrol(1)'s -T flag. The TEKEN_CONS25 kernel configuration
635 options can be used to change the compile-time default back to
638 To prevent graphical artifacts, make sure the TERM environment
639 variable is set to match the terminal emulation that is being
640 performed by syscons(4).
643 The layout of the structure ieee80211req_scan_result has changed.
644 Applications that require wireless scan results (e.g. ifconfig(8))
645 from net80211 need to be recompiled.
647 Applications such as wpa_supplicant(8) may require a full world
648 build without using NO_CLEAN in order to get synchronized with the
652 The iwn(4) driver has been updated to support the 5000 and 5150 series.
653 There's one kernel module for each firmware. Adding "device iwnfw"
654 to the kernel configuration file means including all three firmware
655 images inside the kernel. If you want to include just the one for
656 your wireless card, use the devices iwn4965fw, iwn5000fw or
660 The rc.d/network_ipv6, IPv6 configuration script has been integrated
661 into rc.d/netif. The changes are the following:
663 1. To use IPv6, simply define $ifconfig_IF_ipv6 like $ifconfig_IF
664 for IPv4. For aliases, $ifconfig_IF_aliasN should be used.
665 Note that both variables need the "inet6" keyword at the head.
667 Do not set $ipv6_network_interfaces manually if you do not
668 understand what you are doing. It is not needed in most cases.
670 $ipv6_ifconfig_IF and $ipv6_ifconfig_IF_aliasN still work, but
673 2. $ipv6_enable is obsolete. Use $ipv6_prefer and
674 "inet6 accept_rtadv" keyword in ifconfig(8) instead.
676 If you define $ipv6_enable=YES, it means $ipv6_prefer=YES and
677 all configured interfaces have "inet6 accept_rtadv" in the
678 $ifconfig_IF_ipv6. These are for backward compatibility.
680 3. A new variable $ipv6_prefer has been added. If NO, IPv6
681 functionality of interfaces with no corresponding
682 $ifconfig_IF_ipv6 is disabled by using "inet6 ifdisabled" flag,
683 and the default address selection policy of ip6addrctl(8)
684 is the IPv4-preferred one (see rc.d/ip6addrctl for more details).
685 Note that if you want to configure IPv6 functionality on the
686 disabled interfaces after boot, first you need to clear the flag by
687 using ifconfig(8) like:
689 ifconfig em0 inet6 -ifdisabled
691 If YES, the default address selection policy is set as
694 The default value of $ipv6_prefer is NO.
696 4. If your system need to receive Router Advertisement messages,
697 define "inet6 accept_rtadv" in $ifconfig_IF_ipv6. The rc(8)
698 scripts automatically invoke rtsol(8) when the interface becomes
699 UP. The Router Advertisement messages are used for SLAAC
700 (State-Less Address AutoConfiguration).
703 802.11s D3.03 support was committed. This is incompatible with the
704 previous code, which was based on D3.0.
707 A sysctl variable net.inet6.ip6.accept_rtadv now sets the default value
708 of a per-interface flag ND6_IFF_ACCEPT_RTADV, not a global knob to
709 control whether accepting Router Advertisement messages or not.
710 Also, a per-interface flag ND6_IFF_AUTO_LINKLOCAL has been added and
711 a sysctl variable net.inet6.ip6.auto_linklocal is its default value.
712 The ifconfig(8) utility now supports these flags.
715 ZFS snapshots are now mounted with MNT_IGNORE flag. Use -v option for
716 mount(8) and -a option for df(1) to see them.
719 The old tunable hw.bus.devctl_disable has been superseded by
720 hw.bus.devctl_queue. hw.bus.devctl_disable=1 in loader.conf should be
721 replaced by hw.bus.devctl_queue=0. The default for this new tunable
725 Remove the option STOP_NMI. The default action is now to use NMI only
726 for KDB via the newly introduced function stop_cpus_hard() and
727 maintain stop_cpus() to just use a normal IPI_STOP on ia32 and amd64.
730 The stable/8 branch created in subversion. This corresponds to the
731 RELENG_8 branch in CVS.
734 Bump the shared library version numbers for all libraries that do not
735 use symbol versioning as part of the 8.0-RELEASE cycle. Bump
736 __FreeBSD_version to 800105.
739 Due to changes in the implementation of virtual network stack support,
740 all network-related kernel modules must be recompiled. As this change
741 breaks the ABI, bump __FreeBSD_version to 800104.
744 The TOE interface to the TCP syncache has been modified to remove
745 struct tcpopt (<netinet/tcp_var.h>) from the ABI of the network stack.
746 The cxgb driver is the only TOE consumer affected by this change, and
747 needs to be recompiled along with the kernel. As this change breaks
748 the ABI, bump __FreeBSD_version to 800103.
751 Padding has been added to struct tcpcb, sackhint and tcpstat in
752 <netinet/tcp_var.h> to facilitate future MFCs and bug fixes whilst
753 maintaining the ABI. However, this change breaks the ABI, so bump
754 __FreeBSD_version to 800102. User space tools that rely on the size of
755 any of these structs (e.g. sockstat) need to be recompiled.
758 The NFS_LEGACYRPC option has been removed along with the old kernel
759 RPC implementation that this option selected. Kernel configurations
760 may need to be adjusted.
763 The network interface device nodes at /dev/net/<interface> have been
764 removed. All ioctl operations can be performed the normal way using
765 routing sockets. The kqueue functionality can generally be replaced
766 with routing sockets.
769 The documentation from the FreeBSD Documentation Project (Handbook,
770 FAQ, etc.) is now installed via packages by sysinstall(8) and under
771 the /usr/local/share/doc/freebsd directory instead of /usr/share/doc.
774 The ABI of various structures related to the SYSV IPC API have been
775 changed. As a result, the COMPAT_FREEBSD[456] and COMPAT_43 kernel
776 options now all require COMPAT_FREEBSD7. Bump __FreeBSD_version to
780 Layout of struct vnet has changed as routing related variables were
781 moved to their own Vimage module. Modules need to be recompiled. Bump
782 __FreeBSD_version to 800099.
785 NGROUPS_MAX and NGROUPS have been increased from 16 to 1023 and 1024
786 respectively. As long as no more than 16 groups per process are used,
787 no changes should be visible. When more than 16 groups are used, old
788 binaries may fail if they call getgroups() or getgrouplist() with
789 statically sized storage. Recompiling will work around this, but
790 applications should be modified to use dynamically allocated storage
791 for group arrays as POSIX.1-2008 does not cap an implementation's
792 number of supported groups at NGROUPS_MAX+1 as previous versions did.
794 NFS and portalfs mounts may also be affected as the list of groups is
795 truncated to 16. Users of NFS who use more than 16 groups, should
796 take care that negative group permissions are not used on the exported
797 file systems as they will not be reliable unless a GSSAPI based
798 authentication method is used.
801 The compiling option ADAPTIVE_LOCKMGRS has been introduced. This
802 option compiles in the support for adaptive spinning for lockmgrs
803 which want to enable it. The lockinit() function now accepts the flag
804 LK_ADAPTIVE in order to make the lock object subject to adaptive
805 spinning when both held in write and read mode.
808 The layout of the structure returned by IEEE80211_IOC_STA_INFO has
809 changed. User applications that use this ioctl need to be rebuilt.
812 The layout of struct thread has changed. Kernel and modules need to
816 The layout of structs ifnet, domain, protosw and vnet_net has changed.
817 Kernel modules need to be rebuilt. Bump __FreeBSD_version to 800097.
820 window(1) has been removed from the base system. It can now be
821 installed from ports. The port is called misc/window.
824 The way we are storing and accessing `routing table' entries has
825 changed. Programs reading the FIB, like netstat, need to be
829 A new netisr implementation has been added for FreeBSD 8. Network
830 file system modules, such as igmp, ipdivert, and others, should be
832 Bump __FreeBSD_version to 800096.
835 Remove the tunable/sysctl debug.mpsafevfs as its initial purpose is no
839 Add VOP_ACCESSX(9). File system modules need to be rebuilt.
840 Bump __FreeBSD_version to 800094.
843 Add mnt_xflag field to 'struct mount'. File system modules need to be
845 Bump __FreeBSD_version to 800093.
848 The compiling option ADAPTIVE_SX has been retired while it has been
849 introduced the option NO_ADAPTIVE_SX which handles the reversed logic.
850 The KPI for sx_init_flags() changes as accepting flags:
851 SX_ADAPTIVESPIN flag has been retired while the SX_NOADAPTIVE flag has
852 been introduced in order to handle the reversed logic.
853 Bump __FreeBSD_version to 800092.
856 Add support for hierarchical jails. Remove global securelevel.
857 Bump __FreeBSD_version to 800091.
860 The layout of struct vnet_net has changed, therefore modules
862 Bump __FreeBSD_version to 800090.
865 The newly imported zic(8) produces a new format in the output. Please
866 run tzsetup(8) to install the newly created data to /etc/localtime.
869 The sysctl tree for the usb stack has renamed from hw.usb2.* to
870 hw.usb.* and is now consistent again with previous releases.
873 802.11 monitor mode support was revised and driver api's were changed.
874 Drivers dependent on net80211 now support DLT_IEEE802_11_RADIO instead
875 of DLT_IEEE802_11. No user-visible data structures were changed but
876 applications that use DLT_IEEE802_11 may require changes.
877 Bump __FreeBSD_version to 800088.
880 The layout of the following structs has changed: sysctl_oid,
881 socket, ifnet, inpcbinfo, tcpcb, syncache_head, vnet_inet,
882 vnet_inet6 and vnet_ipfw. Most modules need to be rebuild or
883 panics may be experienced. World rebuild is required for
884 correctly checking networking state from userland.
885 Bump __FreeBSD_version to 800085.
888 MLDv2 and Source-Specific Multicast (SSM) have been merged
889 to the IPv6 stack. VIMAGE hooks are in but not yet used.
890 The implementation of SSM within FreeBSD's IPv6 stack closely
891 follows the IPv4 implementation.
893 For kernel developers:
895 * The most important changes are that the ip6_output() and
896 ip6_input() paths no longer take the IN6_MULTI_LOCK,
897 and this lock has been downgraded to a non-recursive mutex.
899 * As with the changes to the IPv4 stack to support SSM, filtering
900 of inbound multicast traffic must now be performed by transport
901 protocols within the IPv6 stack. This does not apply to TCP and
902 SCTP, however, it does apply to UDP in IPv6 and raw IPv6.
904 * The KPIs used by IPv6 multicast are similar to those used by
905 the IPv4 stack, with the following differences:
906 * im6o_mc_filter() is analogous to imo_multicast_filter().
907 * The legacy KAME entry points in6_joingroup and in6_leavegroup()
908 are shimmed to in6_mc_join() and in6_mc_leave() respectively.
909 * IN6_LOOKUP_MULTI() has been deprecated and removed.
910 * IPv6 relies on MLD for the DAD mechanism. KAME's internal KPIs
911 for MLDv1 have an additional 'timer' argument which is used to
912 jitter the initial membership report for the solicited-node
913 multicast membership on-link.
914 * This is not strictly needed for MLDv2, which already jitters
915 its report transmissions. However, the 'timer' argument is
916 preserved in case MLDv1 is active on the interface.
918 * The KAME linked-list based IPv6 membership implementation has
919 been refactored to use a vector similar to that used by the IPv4
921 Code which maintains a list of its own multicast memberships
922 internally, e.g. carp, has been updated to reflect the new
925 * There is a known Lock Order Reversal (LOR) due to in6_setscope()
926 acquiring the IF_AFDATA_LOCK and being called within ip6_output().
927 Whilst MLDv2 tries to avoid this otherwise benign LOR, it is an
928 implementation constraint which needs to be addressed in HEAD.
930 For application developers:
932 * The changes are broadly similar to those made for the IPv4
935 * The use of IPv4 and IPv6 multicast socket options on the same
936 socket, using mapped addresses, HAS NOT been tested or supported.
938 * There are a number of issues with the implementation of various
939 IPv6 multicast APIs which need to be resolved in the API surface
940 before the implementation is fully compatible with KAME userland
941 use, and these are mostly to do with interface index treatment.
943 * The literature available discusses the use of either the delta / ASM
944 API with setsockopt(2)/getsockopt(2), or the full-state / ASM API
945 using setsourcefilter(3)/getsourcefilter(3). For more information
946 please refer to RFC 3768, 'Socket Interface Extensions for
947 Multicast Source Filters'.
949 * Applications which use the published RFC 3678 APIs should be fine.
951 For systems administrators:
953 * The mtest(8) utility has been refactored to support IPv6, in
954 addition to IPv4. Interface addresses are no longer accepted
955 as arguments, their names must be used instead. The utility
956 will map the interface name to its first IPv4 address as
957 returned by getifaddrs(3).
959 * The ifmcstat(8) utility has also been updated to print the MLDv2
960 endpoint state and source filter lists via sysctl(3).
962 * The net.inet6.ip6.mcast.loop sysctl may be tuned to 0 to disable
963 loopback of IPv6 multicast datagrams by default; it defaults to 1
964 to preserve the existing behaviour. Disabling multicast loopback is
965 recommended for optimal system performance.
967 * The IPv6 MROUTING code has been changed to examine this sysctl
968 instead of attempting to perform a group lookup before looping
969 back forwarded datagrams.
971 Bump __FreeBSD_version to 800084.
974 Implement low-level Bluetooth HCI API.
975 Bump __FreeBSD_version to 800083.
978 The layout of struct malloc_type, used by modules to register new
979 memory allocation types, has changed. Most modules will need to
980 be rebuilt or panics may be experienced.
981 Bump __FreeBSD_version to 800081.
984 Anticipate overflowing inp_flags - add inp_flags2.
985 This changes most offsets in inpcb, so checking v4 connection
986 state will require a world rebuild.
987 Bump __FreeBSD_version to 800080.
990 Add an llentry to struct route and struct route_in6. Modules
991 embedding a struct route will need to be recompiled.
992 Bump __FreeBSD_version to 800079.
995 The size of rt_metrics_lite and by extension rtentry has changed.
996 Networking administration apps will need to be recompiled.
997 The route command now supports show as an alias for get, weighting
998 of routes, sticky and nostick flags to alter the behavior of stateful
1000 Bump __FreeBSD_version to 800078.
1003 Do not use Giant for kbdmux(4) locking. This is wrong and
1004 apparently causing more problems than it solves. This will
1005 re-open the issue where interrupt handlers may race with
1006 kbdmux(4) in polling mode. Typical symptoms include (but
1007 not limited to) duplicated and/or missing characters when
1008 low level console functions (such as gets) are used while
1009 interrupts are enabled (for example geli password prompt,
1010 mountroot prompt etc.). Disabling kbdmux(4) may help.
1013 The size of structs vnet_net, vnet_inet and vnet_ipfw has changed;
1014 kernel modules referencing any of the above need to be recompiled.
1015 Bump __FreeBSD_version to 800075.
1018 GEOM_PART has become the default partition slicer for storage devices,
1019 replacing GEOM_MBR, GEOM_BSD, GEOM_PC98 and GEOM_GPT slicers. It
1020 introduces some changes:
1022 MSDOS/EBR: the devices created from MSDOS extended partition entries
1023 (EBR) can be named differently than with GEOM_MBR and are now symlinks
1024 to devices with offset-based names. fstabs may need to be modified.
1026 BSD: the "geometry does not match label" warning is harmless in most
1027 cases but it points to problems in file system misalignment with
1028 disk geometry. The "c" partition is now implicit, covers the whole
1029 top-level drive and cannot be (mis)used by users.
1031 General: Kernel dumps are now not allowed to be written to devices
1032 whose partition types indicate they are meant to be used for file
1033 systems (or, in case of MSDOS partitions, as something else than
1036 Most of these changes date approximately from 200812.
1039 The uscanner(4) driver has been removed from the kernel. This follows
1040 Linux removing theirs in 2.6 and making libusb the default interface
1041 (supported by sane).
1044 The multicast forwarding code has been cleaned up. netstat(1)
1045 only relies on KVM now for printing bandwidth upcall meters.
1046 The IPv4 and IPv6 modules are split into ip_mroute_mod and
1047 ip6_mroute_mod respectively. The config(5) options for statically
1048 compiling this code remain the same, i.e. 'options MROUTING'.
1051 Support for the IFF_NEEDSGIANT network interface flag has been
1052 removed, which means that non-MPSAFE network device drivers are no
1053 longer supported. In particular, if_ar, if_sr, and network device
1054 drivers from the old (legacy) USB stack can no longer be built or
1058 POSIX.1 Native Language Support (NLS) has been enabled in libc and
1059 a bunch of new language catalog files have also been added.
1060 This means that some common libc messages are now localized and
1061 they depend on the LC_MESSAGES environmental variable.
1064 The k8temp(4) driver has been renamed to amdtemp(4) since
1065 support for Family 10 and Family 11 CPU families was added.
1068 IGMPv3 and Source-Specific Multicast (SSM) have been merged
1069 to the IPv4 stack. VIMAGE hooks are in but not yet used.
1071 For kernel developers, the most important changes are that the
1072 ip_output() and ip_input() paths no longer take the IN_MULTI_LOCK(),
1073 and this lock has been downgraded to a non-recursive mutex.
1075 Transport protocols (UDP, Raw IP) are now responsible for filtering
1076 inbound multicast traffic according to group membership and source
1077 filters. The imo_multicast_filter() KPI exists for this purpose.
1078 Transports which do not use multicast (SCTP, TCP) already reject
1079 multicast by default. Forwarding and receive performance may improve
1080 as a mutex acquisition is no longer needed in the ip_input()
1081 low-level input path. in_addmulti() and in_delmulti() are shimmed
1082 to new KPIs which exist to support SSM in-kernel.
1084 For application developers, it is recommended that loopback of
1085 multicast datagrams be disabled for best performance, as this
1086 will still cause the lock to be taken for each looped-back
1087 datagram transmission. The net.inet.ip.mcast.loop sysctl may
1088 be tuned to 0 to disable loopback by default; it defaults to 1
1089 to preserve the existing behaviour.
1091 For systems administrators, to obtain best performance with
1092 multicast reception and multiple groups, it is always recommended
1093 that a card with a suitably precise hash filter is used. Hash
1094 collisions will still result in the lock being taken within the
1095 transport protocol input path to check group membership.
1097 If deploying FreeBSD in an environment with IGMP snooping switches,
1098 it is recommended that the net.inet.igmp.sendlocal sysctl remain
1099 enabled; this forces 224.0.0.0/24 group membership to be announced
1102 The size of 'struct igmpstat' has changed; netstat needs to be
1103 recompiled to reflect this.
1104 Bump __FreeBSD_version to 800070.
1107 libusb20.so.1 is now installed as libusb.so.1 and the ports system
1108 updated to use it. This requires a buildworld/installworld in order to
1109 update the library and dependencies (usbconfig, etc). Its advisable to
1110 rebuild all ports which uses libusb. More specific directions are given
1111 in the ports collection UPDATING file. Any /etc/libmap.conf entries for
1112 libusb are no longer required and can be removed.
1115 A workaround is committed to allow the creation of System V shared
1116 memory segment of size > 2 GB on the 64-bit architectures.
1117 Due to a limitation of the existing ABI, the shm_segsz member
1118 of the struct shmid_ds, returned by shmctl(IPC_STAT) call is
1119 wrong for large segments. Note that limits must be explicitly
1120 raised to allow such segments to be created.
1123 The layout of struct ifnet has changed, requiring a rebuild of all
1124 network device driver modules.
1127 The /dev handling for the new USB stack has changed, a
1128 buildworld/installworld is required for libusb20.
1131 The new USB2 stack has now been permanently moved in and all kernel and
1132 module names reverted to their previous values (eg, usb, ehci, ohci,
1133 ums, ...). The old usb stack can be compiled in by prefixing the name
1134 with the letter 'o', the old usb modules have been removed.
1135 Updating entry 20090216 for xorg and 20090215 for libmap may still
1139 The rc.conf(5) option if_up_delay has been renamed to
1140 defaultroute_delay to better reflect its purpose. If you have
1141 customized this setting in /etc/rc.conf you need to update it to
1145 xorg 7.4 wants to configure its input devices via hald which does not
1146 yet work with USB2. If the keyboard/mouse does not work in xorg then
1148 Option "AllowEmptyInput" "off"
1149 to your ServerLayout section. This will cause X to use the configured
1150 kbd and mouse sections from your xorg.conf.
1153 The GENERIC kernels for all architectures now default to the new USB2
1154 stack. No kernel config options or code have been removed so if a
1155 problem arises please report it and optionally revert to the old USB
1156 stack. If you are loading USB kernel modules or have a custom kernel
1157 that includes GENERIC then ensure that usb names are also changed over,
1158 eg uftdi -> usb2_serial_ftdi.
1160 Older programs linked against the ports libusb 0.1 need to be
1161 redirected to the new stack's libusb20. /etc/libmap.conf can
1163 # Map old usb library to new one for usb2 stack
1164 libusb-0.1.so.8 libusb20.so.1
1167 All USB ethernet devices now attach as interfaces under the name ueN
1168 (eg. ue0). This is to provide a predictable name as vendors often
1169 change usb chipsets in a product without notice.
1172 The ichsmb(4) driver has been changed to require SMBus slave
1173 addresses be left-justified (xxxxxxx0b) rather than right-justified.
1174 All of the other SMBus controller drivers require left-justified
1175 slave addresses, so this change makes all the drivers provide the
1179 INET6 statistics (struct ip6stat) was updated.
1180 netstat(1) needs to be recompiled.
1183 NTFS has been removed from GENERIC kernel on amd64 to match
1184 GENERIC on i386. Should not cause any issues since mount_ntfs(8)
1185 will load ntfs.ko module automatically when NTFS support is
1186 actually needed, unless ntfs.ko is not installed or security
1187 level prohibits loading kernel modules. If either is the case,
1188 "options NTFS" has to be added into kernel config.
1191 TCP Appropriate Byte Counting (RFC 3465) support added to kernel.
1192 New field in struct tcpcb breaks ABI, so bump __FreeBSD_version to
1193 800061. User space tools that rely on the size of struct tcpcb in
1194 tcp_var.h (e.g. sockstat) need to be recompiled.
1197 ng_tty(4) module updated to match the new TTY subsystem.
1198 Due to API change, user-level applications must be updated.
1199 New API support added to mpd5 CVS and expected to be present
1200 in next mpd5.3 release.
1203 With __FreeBSD_version 800060 the makefs tool is part of
1204 the base system (it was a port).
1207 The afdata and ifnet locks have been changed from mutexes to
1208 rwlocks, network modules will need to be re-compiled.
1211 __FreeBSD_version 800059 incorporates the new arp-v2 rewrite.
1212 RTF_CLONING, RTF_LLINFO and RTF_WASCLONED flags are eliminated.
1213 The new code reduced struct rtentry{} by 16 bytes on 32-bit
1214 architecture and 40 bytes on 64-bit architecture. The userland
1215 applications "arp" and "ndp" have been updated accordingly.
1216 The output from "netstat -r" shows only routing entries and
1217 none of the L2 information.
1220 __FreeBSD_version 800057 marks the switchover from the
1221 binary ath hal to source code. Users must add the line:
1223 options AH_SUPPORT_AR5416
1225 to their kernel config files when specifying:
1229 The ath_hal module no longer exists; the code is now compiled
1230 together with the driver in the ath module. It is now
1231 possible to tailor chip support (i.e. reduce the set of chips
1232 and thereby the code size); consult ath_hal(4) for details.
1235 __FreeBSD_version 800054 adds memory barriers to
1236 <machine/atomic.h>, new interfaces to ifnet to facilitate
1237 multiple hardware transmit queues for cards that support
1238 them, and a lock-less ring-buffer implementation to
1239 enable drivers to more efficiently manage queueing of
1243 A new version of ZFS (version 13) has been merged to -HEAD.
1244 This version has zpool attribute "listsnapshots" off by
1245 default, which means "zfs list" does not show snapshots,
1246 and is the same as Solaris behavior.
1249 dummynet(4) ABI has changed. ipfw(8) needs to be recompiled.
1252 The uhci, ohci, ehci and slhci USB Host controller drivers have
1253 been put into separate modules. If you load the usb module
1254 separately through loader.conf you will need to load the
1255 appropriate *hci module as well. E.g. for a UHCI-based USB 2.0
1256 controller add the following to loader.conf:
1262 The ABI used by the PMC toolset has changed. Please keep
1263 userland (libpmc(3)) and the kernel module (hwpmc(4)) in
1267 atapci kernel module now includes only generic PCI ATA
1268 driver. AHCI driver moved to ataahci kernel module.
1269 All vendor-specific code moved into separate kernel modules:
1270 ataacard, ataacerlabs, ataadaptec, ataamd, ataati, atacenatek,
1271 atacypress, atacyrix, atahighpoint, ataintel, ataite, atajmicron,
1272 atamarvell, atamicron, atanational, atanetcell, atanvidia,
1273 atapromise, ataserverworks, atasiliconimage, atasis, atavia
1276 The TTY subsystem of the kernel has been replaced by a new
1277 implementation, which provides better scalability and an
1278 improved driver model. Most common drivers have been migrated to
1279 the new TTY subsystem, while others have not. The following
1280 drivers have not yet been ported to the new TTY layer:
1283 cy, digi, rc, rp, sio
1289 ng_h4, ng_tty, ppp, sl, snp
1291 Adding these drivers to your kernel configuration file shall
1292 cause compilation to fail.
1295 ntpd has been upgraded to 4.2.4p5.
1298 OpenSSH has been upgraded to 5.1p1.
1300 For many years, FreeBSD's version of OpenSSH preferred DSA
1301 over RSA for host and user authentication keys. With this
1302 upgrade, we've switched to the vendor's default of RSA over
1303 DSA. This may cause upgraded clients to warn about unknown
1304 host keys even for previously known hosts. Users should
1305 follow the usual procedure for verifying host keys before
1306 accepting the RSA key.
1308 This can be circumvented by setting the "HostKeyAlgorithms"
1309 option to "ssh-dss,ssh-rsa" in ~/.ssh/config or on the ssh
1312 Please note that the sequence of keys offered for
1313 authentication has been changed as well. You may want to
1314 specify IdentityFile in a different order to revert this
1318 The sio(4) driver has been removed from the i386 and amd64
1319 kernel configuration files. This means uart(4) is now the
1320 default serial port driver on those platforms as well.
1322 To prevent collisions with the sio(4) driver, the uart(4) driver
1323 uses different names for its device nodes. This means the
1324 onboard serial port will now most likely be called "ttyu0"
1325 instead of "ttyd0". You may need to reconfigure applications to
1326 use the new device names.
1328 When using the serial port as a boot console, be sure to update
1329 /boot/device.hints and /etc/ttys before booting the new kernel.
1330 If you forget to do so, you can still manually specify the hints
1331 at the loader prompt:
1333 set hint.uart.0.at="isa"
1334 set hint.uart.0.port="0x3F8"
1335 set hint.uart.0.flags="0x10"
1336 set hint.uart.0.irq="4"
1340 The gpt(8) utility has been removed. Use gpart(8) to partition
1344 The version that Linuxulator emulates was changed from 2.4.2
1345 to 2.6.16. If you experience any problems with Linux binaries
1346 please try to set sysctl compat.linux.osrelease to 2.4.2 and
1347 if it fixes the problem contact emulation mailing list.
1350 ISDN4BSD (I4B) was removed from the src tree. You may need to
1351 update a your kernel configuration and remove relevant entries.
1354 I have checked in code to support multiple routing tables.
1355 See the man pages setfib(1) and setfib(2).
1356 This is a hopefully backwards compatible version,
1357 but to make use of it you need to compile your kernel
1358 with options ROUTETABLES=2 (or more up to 16).
1361 The 802.11 wireless support was redone to enable multi-bss
1362 operation on devices that are capable. The underlying device
1363 is no longer used directly but instead wlanX devices are
1364 cloned with ifconfig. This requires changes to rc.conf files.
1365 For example, change:
1366 ifconfig_ath0="WPA DHCP"
1369 ifconfig_wlan0="WPA DHCP"
1370 see rc.conf(5) for more details. In addition, mergemaster of
1371 /etc/rc.d is highly recommended. Simultaneous update of userland
1372 and kernel wouldn't hurt either.
1374 As part of the multi-bss changes the wlan_scan_ap and wlan_scan_sta
1375 modules were merged into the base wlan module. All references
1376 to these modules (e.g. in kernel config files) must be removed.
1379 psm(4) has gained write(2) support in native operation level.
1380 Arbitrary commands can be written to /dev/psm%d and status can
1381 be read back from it. Therefore, an application is responsible
1382 for status validation and error recovery. It is a no-op in
1383 other operation levels.
1386 Support for KSE threading has been removed from the kernel. To
1387 run legacy applications linked against KSE libmap.conf may
1388 be used. The following libmap.conf may be used to ensure
1389 compatibility with any prior release:
1391 libpthread.so.1 libthr.so.1
1392 libpthread.so.2 libthr.so.2
1393 libkse.so.3 libthr.so.3
1396 The layout of struct vmspace has changed. This affects libkvm
1397 and any executables that link against libkvm and use the
1398 kvm_getprocs() function. In particular, but not exclusively,
1399 it affects ps(1), fstat(1), pkill(1), systat(1), top(1) and w(1).
1400 The effects are minimal, but it's advisable to upgrade world
1404 The latest em driver no longer has support in it for the
1405 82575 adapter, this is now moved to the igb driver. The
1406 split was done to make new features that are incompatible
1407 with older hardware easier to do.
1410 The new geom_lvm(4) geom class has been renamed to geom_linux_lvm(4),
1411 likewise the kernel option is now GEOM_LINUX_LVM.
1414 The default NFS mount mode has changed from UDP to TCP for
1415 increased reliability. If you rely on (insecurely) NFS
1416 mounting across a firewall you may need to update your
1420 Belatedly note the addition of m_collapse for compacting
1424 The fts(3) structures have been changed to use adequate
1425 integer types for their members and so to be able to cope
1426 with huge file trees. The old fts(3) ABI is preserved
1427 through symbol versioning in libc, so third-party binaries
1428 using fts(3) should still work, although they will not take
1429 advantage of the extended types. At the same time, some
1430 third-party software might fail to build after this change
1431 due to unportable assumptions made in its source code about
1432 fts(3) structure members. Such software should be fixed
1433 by its vendor or, in the worst case, in the ports tree.
1434 FreeBSD_version 800015 marks this change for the unlikely
1435 case that a portable fix is impossible.
1438 To upgrade to -current after this date, you must be running
1439 FreeBSD not older than 6.0-RELEASE. Upgrading to -current
1440 from 5.x now requires a stop over at RELENG_6 or RELENG_7 systems.
1443 The ADAPTIVE_GIANT kernel option has been retired because its
1444 functionality is the default now.
1447 The AT keyboard emulation of sunkbd(4) has been turned on
1448 by default. In order to make the special symbols of the Sun
1449 keyboards driven by sunkbd(4) work under X these now have
1450 to be configured the same way as Sun USB keyboards driven
1451 by ukbd(4) (which also does AT keyboard emulation), f.e.:
1453 Option "XkbLayout" "us"
1454 Option "XkbRules" "xorg"
1455 Option "XkbSymbols" "pc(pc105)+sun_vndr/usb(sun_usb)+us"
1458 It has been decided that it is desirable to provide ABI
1459 backwards compatibility to the FreeBSD 4/5/6 versions of the
1460 PCIOCGETCONF, PCIOCREAD and PCIOCWRITE IOCTLs, which was
1461 broken with the introduction of PCI domain support (see the
1462 20070930 entry). Unfortunately, this required the ABI of
1463 PCIOCGETCONF to be broken again in order to be able to
1464 provide backwards compatibility to the old version of that
1465 IOCTL. Thus consumers of PCIOCGETCONF have to be recompiled
1466 again. As for prominent ports this affects neither pciutils
1467 nor xorg-server this time, the hal port needs to be rebuilt
1471 The misnamed kthread_create() and friends have been renamed
1472 to kproc_create() etc. Many of the callers already
1473 used kproc_start()..
1474 I will return kthread_create() and friends in a while
1475 with implementations that actually create threads, not procs.
1476 Renaming corresponds with version 800002.
1485 Avoid using make -j when upgrading. While generally safe, there are
1486 sometimes problems using -j to upgrade. If your upgrade fails with
1487 -j, please try again without -j. From time to time in the past there
1488 have been problems using -j with buildworld and/or installworld. This
1489 is especially true when upgrading between "distant" versions (eg one
1490 that cross a major release boundary or several minor releases, or when
1491 several months have passed on the -current branch).
1493 Sometimes, obscure build problems are the result of environment
1494 poisoning. This can happen because the make utility reads its
1495 environment when searching for values for global variables. To run
1496 your build attempts in an "environmental clean room", prefix all make
1497 commands with 'env -i '. See the env(1) manual page for more details.
1499 When upgrading from one major version to another it is generally best
1500 to upgrade to the latest code in the currently installed branch first,
1501 then do an upgrade to the new branch. This is the best-tested upgrade
1502 path, and has the highest probability of being successful. Please try
1503 this approach before reporting problems with a major version upgrade.
1507 When upgrading the boot ZFS pool to a new version, always follow
1510 1.) recompile and reinstall the ZFS boot loader and boot block
1511 (this is part of "make buildworld" and "make installworld")
1513 2.) update the ZFS boot block on your boot drive
1515 The following example updates the ZFS boot block on the first
1516 partition (freebsd-boot) of a GPT partitioned drive ad0:
1517 "gpart bootcode -p /boot/gptzfsboot -i 1 ad0"
1519 Non-boot pools do not need these updates.
1523 If you are updating from a prior version of FreeBSD (even one just
1524 a few days old), you should follow this procedure. It is the most
1525 failsafe as it uses a /usr/obj tree with a fresh mini-buildworld,
1527 make kernel-toolchain
1528 make -DALWAYS_CHECK_MAKE buildkernel KERNCONF=YOUR_KERNEL_HERE
1529 make -DALWAYS_CHECK_MAKE installkernel KERNCONF=YOUR_KERNEL_HERE
1531 To test a kernel once
1532 ---------------------
1533 If you just want to boot a kernel once (because you are not sure
1534 if it works, or if you want to boot a known bad kernel to provide
1535 debugging information) run
1536 make installkernel KERNCONF=YOUR_KERNEL_HERE KODIR=/boot/testkernel
1537 nextboot -k testkernel
1539 To just build a kernel when you know that it won't mess you up
1540 --------------------------------------------------------------
1541 This assumes you are already running a CURRENT system. Replace
1542 ${arch} with the architecture of your machine (e.g. "i386",
1543 "arm", "amd64", "ia64", "pc98", "sparc64", "powerpc", "mips", etc).
1545 cd src/sys/${arch}/conf
1546 config KERNEL_NAME_HERE
1547 cd ../compile/KERNEL_NAME_HERE
1552 If this fails, go to the "To build a kernel" section.
1554 To rebuild everything and install it on the current system.
1555 -----------------------------------------------------------
1556 # Note: sometimes if you are running current you gotta do more than
1557 # is listed here if you are upgrading from a really old current.
1559 <make sure you have good level 0 dumps>
1561 make kernel KERNCONF=YOUR_KERNEL_HERE
1563 <reboot in single user> [3]
1571 To cross-install current onto a separate partition
1572 --------------------------------------------------
1573 # In this approach we use a separate partition to hold
1574 # current's root, 'usr', and 'var' directories. A partition
1575 # holding "/", "/usr" and "/var" should be about 2GB in
1578 <make sure you have good level 0 dumps>
1581 make buildkernel KERNCONF=YOUR_KERNEL_HERE
1582 <maybe newfs current's root partition>
1583 <mount current's root partition on directory ${CURRENT_ROOT}>
1584 make installworld DESTDIR=${CURRENT_ROOT}
1585 make distribution DESTDIR=${CURRENT_ROOT} # if newfs'd
1586 make installkernel KERNCONF=YOUR_KERNEL_HERE DESTDIR=${CURRENT_ROOT}
1587 cp /etc/fstab ${CURRENT_ROOT}/etc/fstab # if newfs'd
1588 <edit ${CURRENT_ROOT}/etc/fstab to mount "/" from the correct partition>
1589 <reboot into current>
1590 <do a "native" rebuild/install as described in the previous section>
1591 <maybe install compatibility libraries from ports/misc/compat*>
1595 To upgrade in-place from stable to current
1596 ----------------------------------------------
1597 <make sure you have good level 0 dumps>
1599 make kernel KERNCONF=YOUR_KERNEL_HERE [8]
1601 <reboot in single user> [3]
1608 Make sure that you've read the UPDATING file to understand the
1609 tweaks to various things you need. At this point in the life
1610 cycle of current, things change often and you are on your own
1611 to cope. The defaults can also change, so please read ALL of
1612 the UPDATING entries.
1614 Also, if you are tracking -current, you must be subscribed to
1615 freebsd-current@freebsd.org. Make sure that before you update
1616 your sources that you have read and understood all the recent
1617 messages there. If in doubt, please track -stable which has
1618 much fewer pitfalls.
1620 [1] If you have third party modules, such as vmware, you
1621 should disable them at this point so they don't crash your
1624 [3] From the bootblocks, boot -s, and then do
1629 adjkerntz -i # if CMOS is wall time
1630 Also, when doing a major release upgrade, it is required that
1631 you boot into single user mode to do the installworld.
1633 [4] Note: This step is non-optional. Failure to do this step
1634 can result in a significant reduction in the functionality of the
1635 system. Attempting to do it by hand is not recommended and those
1636 that pursue this avenue should read this file carefully, as well
1637 as the archives of freebsd-current and freebsd-hackers mailing lists
1638 for potential gotchas. The -U option is also useful to consider.
1639 See mergemaster(8) for more information.
1641 [5] Usually this step is a noop. However, from time to time
1642 you may need to do this if you get unknown user in the following
1643 step. It never hurts to do it all the time. You may need to
1644 install a new mergemaster (cd src/usr.sbin/mergemaster && make
1645 install) after the buildworld before this step if you last updated
1646 from [78]-stable or 9-stable before 20130430.
1648 [6] This only deletes old files and directories. Old libraries
1649 can be deleted by "make delete-old-libs", but you have to make
1650 sure that no program is using those libraries anymore.
1652 [8] In order to have a kernel that can run the 4.x binaries needed to
1653 do an installworld, you must include the COMPAT_FREEBSD4 option in
1654 your kernel. Failure to do so may leave you with a system that is
1655 hard to boot to recover. A similar kernel option COMPAT_FREEBSD5 is
1656 required to run the 5.x binaries on more recent kernels. And so on
1657 for COMPAT_FREEBSD6 and COMPAT_FREEBSD7.
1659 Make sure that you merge any new devices from GENERIC since the
1660 last time you updated your kernel config file.
1662 [9] When checking out sources, you must include the -P flag to have
1663 cvs prune empty directories.
1665 If CPUTYPE is defined in your /etc/make.conf, make sure to use the
1666 "?=" instead of the "=" assignment operator, so that buildworld can
1667 override the CPUTYPE if it needs to.
1669 MAKEOBJDIRPREFIX must be defined in an environment variable, and
1670 not on the command line, or in /etc/make.conf. buildworld will
1671 warn if it is improperly defined.
1674 This file contains a list, in reverse chronological order, of major
1675 breakages in tracking -current. It is not guaranteed to be a complete
1676 list of such breakages, and only contains entries since October 10, 2007.
1677 If you need to see UPDATING entries from before that date, you will need
1678 to fetch an UPDATING file from an older FreeBSD release.
1680 Copyright information:
1682 Copyright 1998-2009 M. Warner Losh. All Rights Reserved.
1684 Redistribution, publication, translation and use, with or without
1685 modification, in full or in part, in any form or format of this
1686 document are permitted without further permission from the author.
1688 THIS DOCUMENT IS PROVIDED BY WARNER LOSH ``AS IS'' AND ANY EXPRESS OR
1689 IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED
1690 WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE
1691 DISCLAIMED. IN NO EVENT SHALL WARNER LOSH BE LIABLE FOR ANY DIRECT,
1692 INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES
1693 (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR
1694 SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
1695 HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT,
1696 STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING
1697 IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE
1698 POSSIBILITY OF SUCH DAMAGE.
1700 Contact Warner Losh if you have any questions about your use of